Re: [datetime] conversion operators are all wrong
Howard Hinnant ha scritto:
In article <AcJuj.17987$FR.94174@twister1.libero.it>,
Alberto Ganesh Barbati <AlbertoBarbati@libero.it> wrote:
I see. I didn't realize that the conversion from tick_type to duration
type was implicit. I have blindly assumed it be explicit and I am
surprised to know that it's not so, given that it would make the
conversion much less error-prone. <snip>
This aspect was specifically questioned/considered. Jeff Garland
(author of boost::datetime) successfully argued against making these
constructors explicit on the grounds that it would hurt usability more
than it would help it. This is based on his years of experience with
boost::datetime.
He also pointed out that (in the current WP) when one is accepting a
generalized duration as in:
template <Duration D>
void
f(D d);
(using concepts syntax to restrict the template)
then an integral argument won't bind in your use case anyway. The same
effect will be true if N2498 is accepted (and won't require
concept/constraining):
template <class TT, long long TPS, long long SPT>
void
f(std::duration<TT, TPS, SPT> d);
f(10); // no match
But one can still:
std::seconds s = 5;
which is very convenient.
Thanks a lot. You have been very clear. Now I understand and I totally
agree.
The duration_cast looks definitely a good tool to me. I hope it will be
included. In addition to its intrinsic value, all the conversion
operators could easily be rewritten, for example, as:
Returns: duration_cast<minutes>(*this)
which looks very nice to me as it is both more compact and more readable.
<nod> Thanks. This is what the proposed wording in N2498 looks like.
I just read N2498 and I find it very well thought and better than the
current draft wording. I hope the whole paper (and not only
duration_cast) could be adopted.
HTH,
Ganesh
--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]